Running a tiered link audit

Lesson 16 of 50Module 3 · Reading the link graph6 min read

By Bill Hartzer · Hartzer Consulting · 25+ years in search, domains and expert witness work

What the audit is for

An audit that produces a list of observations is not an audit. Before you open a tool, decide which of three questions you are answering, because they call for different work:

• Was this profile earned or assembled? The due-diligence question. Asked before an acquisition, when taking over an account, or when a competitor's numbers do not match their visibility.
• What is recoverable? The opportunity question, and usually the one that pays. Lost links, dead own-properties, unlinked co-citations.
• Where is the exposure? The risk question. What was placed, what could be identified as placed, and what technical faults are wasting equity that already arrived.

Most commercial audits are sold as the third and deliver the first badly. My experience is that the second is where the value nearly always sits, because a profile with history usually has more equity already earned and mislaid than any campaign will produce in a quarter.

The sequence

Run it in this order. Each step feeds the next, and skipping the early ones makes the later ones uninterpretable.

Steps 1 to 5 take about half an hour once you have done it twice. Steps 6 to 9 are the analytical work and take as long as the profile deserves. Steps 11 and 12 are where the recommendations come from.

Scoring what you found

A rubric will not tell you the answer, but it stops you writing a confident conclusion from three impressions. Score each dimension from 1 to 5, record the evidence beside each score, and refuse to score anything you did not capture.

• Tier-1 quality. Distribution across Trust Flow bands and how many links are editorial rather than directory placements.
• Tier-2 and tier-3 structure. Links per domain by tier. Penalize exactly 1.0 and anything above roughly 3.
• Topical coherence. Do the highest-TTF topics match the subject, and do they rest on enough domains to mean anything?
• Anchor naturalness. Brand and URL anchors leading, with commercial exact match rare. Read the referring-domain column, not the link column.
• Neighborhood. Does Related Sites return recognizable peers or a naming farm?
• Concentration. How much of the profile depends on one or two referrers.
• Durability. Share of referring domains still live, and the Trust Flow distribution of those that stopped.

Two rules make the scores honest. Weight the middle bands heavily — anyone can have a strong outlier and everyone has junk, so the shape of the ordinary majority is the profile. And score tier 1 separately from everything deeper, because a low score on tier 1 is a finding about the site owner and a low score at tier 3 is a finding about the internet.

A worked pass, in miniature

Run the sequence over the two profiles in this module and it resolves in under an hour.

The abandoned domain. Fresh 182 links from 30 domains against Historic 31,004 from 4,180 — under 1 percent of the historic domain count still visible, so the profile is dead rather than merely uncrawled. Tier 1 is exactly 1.0 links per domain across all 30, which is purchase. Tier 3 is 4.5 across 8 domains, which is a link wheel. Anchors are #1 website promotion and seo and one domain producing 123 links on search engine optimization. Related Sites is an article-directory farm. Flow Metric History is flat for a decade after one early Citation Flow spike that decayed away. Clique Hunter against an unrelated industry site returns one shared referrer, and it is Blogspot. Assembled, cheaply, a long time ago, and long since dead.

The earned profile. Fresh 1,130 domains against Historic 22,260 — about 5 percent visible, normal for a large old profile. Links per domain runs 1.35 to 1.96 across tiers with no spike. Anchors are brand-led with no commercial exact match in the top fifteen. Related Sites returns trade and technology press. Ten-year metric history is stable and rising. But the audit also found 68.9 percent of live links arriving through a 301 redirect, 73.5 percent still pointing at http://, 20,768 live links resolving to fetch failures, three service pages that lost more than 20,000 links between them, and an owned property that was once the third-largest referrer and now sends nothing.

That is the shape of a real audit on a healthy site. The link profile is fine; the findings are technical and recoverable, and they are worth more than any acquisition campaign would be that quarter.

What to report

Structure the deliverable in four parts, in this order:

1. What the profile is. Referring domains in both indexes, distribution by Trust Flow band, concentration, and the tier table. Six numbers and one table, not forty screenshots.
2. What you control. Tier 1 findings only. Placed links, exact-match anchors, directory clusters, technical faults on your own side.
3. What you do not control. Tier 2 and deeper, stated plainly as context with an explicit note that no action is available or required.
4. What to do next, ordered by expected value rather than by effort.

Report referring domains everywhere. If a link count appears at all, mark it as distinct or duplicate and say which. Every number carries its index — Fresh or Historic — because that single discipline prevents most of the misreadings this module exists to correct.

What not to report

Four things I would strike from most audits I have been shown:

• A large disavow recommendation. Write it last, and default to nothing. Seven in ten referring domains on an unmanipulated profile are Trust Flow 0; junk arrives because a site exists, not because something went wrong. Disavowing links you did not place, on the strength of a tool's toxicity score, is a way to remove value you cannot get back.
• Anything about tier 2 phrased as a problem. Nobody controls their tier 2. An audit that treats it as a fault is either misunderstanding the tool or manufacturing alarm to sell a cleanup.
• Fresh against Historic. The comparison that makes a healthy profile look catastrophic. If your report contains a decline, verify both numbers came from the same index before it goes anywhere.
• Toxicity scores presented as verdicts. Every one is a vendor's opinion expressed as a number, and none of them has access to how a search engine actually treats a link. Use them to sort, never to decide.

A last note on tone. The most useful audit I ever wrote was about my own site, and it found that I had let an owned property stop linking, left two internal cross-links to rot, and never checked whether my redirect chain was one hop. A report that only flatters is not a report. Neither is one built to frighten. The job is to say what the evidence supports, at the depth the evidence was gathered, and to be clear about which is which.